Institute for ComplexSystems SimulationInstitute for

ComplexSystems Simulation

Understanding the complexity of rural food security:

The use of participatory methods and simulation tools.

Samantha L Dobbie , J. Dyke, K. SchreckenbergInstitute for Complex Systems Simulation (ICSS) Centre for Environmental Sciences (CES) s.dobbie@soton.ac.uk @sldobbie

Institute for ComplexSystems SimulationIntroduction

What is food security?

•“when all people, at all times have physical and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life” FAO – 1996

•The four pillars of food security

Institute for ComplexSystems SimulationMy Research

Motivation

•Understanding how food security can be realised in a developing country context.

•Complex social, ecological & political factors propel food insecurity currently.

Food Security Risk Index 2013

Institute for ComplexSystems SimulationThe use of Agent-based

Modelling • Agents interact within an environment through predisposed rules.

• Patterns at the macro-level emerge as a result of interactions at the local-level.

• Can be abstract, experimental, historical or empirical.

• Offers potential for greater understanding of food security.

Institute for ComplexSystems SimulationManaging Complexity

• How empirical is empirical?

• Stakeholder expectations

• Common criticisms: – Your model is too complex

– Your model is too simple – Your assumptions & parameters are arbitrary

– Your model is a black box

Institute for ComplexSystems Simulation

• How can the challenges of representing reality, technical constraints and meeting the expectations of stakeholders be overcome?

Institute for ComplexSystems SimulationUsing Participatory Methods?

• Can Participatory Rural Appraisals (PRA) be used to parameterise ABM?

• PRA techniques include:

- Matrix scoring

- Seasonal calendars

- Mapping

- Wellbeing ranking

Institute for ComplexSystems SimulationDefining Agent Types

• Using ASSETS wellbeing exercises, 4 agent types were identified:

– Type 1: Male Heads of Household (HH) of medium or rich wellbeing

– Type 2: Male HH of poor or very poor wellbeing

– Type 3: Female HH of medium or rich wellbeing

– Type 4: Female HH of poor or very poor wellbeing

Institute for ComplexSystems SimulationInferring Behavioural Rules

• Participant selection based upon agent types.

• PRA exercise designed to investigate farmer decisions

• Each month a total of 60 counters are split between 15 activities.

• Exercise repeated for a drought year

• Interview style questions explore the impact of input subsidies upon food security.

Institute for ComplexSystems SimulationBuilding a Model

• Results from the PRA exercises used to construct behavioural rules for agent types.

• Model parameterisation also aided by literature.

• Implemented using NetLogo

https://ccl.northwestern.edu/netlogo/

UML Diagram by Stefano Balbi (2013)

Institute for ComplexSystems SimulationDefine-agents ()

• Each of the 15, 808 households to be given an agent type.

• Requires existing household survey data as PRA participants targeted in an un-stratified manner.

• K-means cluster analysis identifies four clusters within the survey data corresponding to four agent types.

• Monte Carlo techniques employed to generate the entire agent population.

  Gender  Male FemaleCluster 1 2 3 4Area of land (ha) 0.9

60.74

0.65

0.22

No. of livestock 4 0 4 0No. of poultry 15 0 15 0Health care 2 2 2 2Food adequacy 2 2 2 1Proportion of sample population (%)

28.9

48.9

 4.4

 1.8

K-means cluster analysi

s

Institute for ComplexSystems SimulationMonte Carlo Techniques

• An empirical cumulative distribution function (CDF) created to determine agent type.

• A random integer between 0 and 100 drawn for each agent and the agent type read from the y-axis.

• Conducted for all 15,808 agents to recreate the empirical distribution.

• Repeated to allocate resource endowments: land area, numbers of livestock, number of poultry & food-adequacy value.

Institute for ComplexSystems SimulationModel Implementation

• Baseline scenario:– 30 % chance of drought, input subsidies available from September to December and accessed by 47 % of the population. Model run for 120 time-steps (10 yrs).

• Drought scenario: – Probability of drought varied between 0 to 100 % in 20% intervals.

• Input subsidy scenario: – Timing of input subsidies varied to be early, typical and late.

Institute for ComplexSystems Simulation

http://www.openabm.org

Institute for ComplexSystems SimulationCritique

• PRA exercise brought greater understanding of the smallholder system and uncovered surprising behaviour.

• A simple yet effective method to parameterise empirical ABM.

• Participatory approach.

• Data limitations – poor availability and reliability.

• Issue of aggregation – inter-village differences not accounted for.

• The need for

Validation -preliminary results are speculative.

Strengths Weaknesses

Institute for ComplexSystems SimulationSummary

• The primary objective: – Can PRA techniques be employed in the parameterisation of empirical agent-based models?

• In this case – yes! – Model implementation allowed complex social, ecological and economic factors to be explored.

• Future work: – Overcoming data limitations – Model validation and verification

Institute for ComplexSystems SimulationAcknowledgements

• This work was supported by an EPSRC Doctoral Training Centre grant (EP/G03690X/1) and took place under the ‘Attaining Sustainable Services from Ecosystems using Trade-off Scenarios’ project (ASSETS; http://espa-assets.org/; NE-J002267-1), funded with support from the United Kingdom’s Ecosystem Services for Poverty Alleviation program (ESPA; www.espa.ac.uk). ESPA receives its funding from the Department for International Development (DFID), the Economic and Social Research Council (ESRC) and the Natural Environment Research Council (NERC).

Institute for ComplexSystems SimulationInstitute for

ComplexSystems Simulation